In order to benefit from the sentinel images for an improved forest management process we developed a road map, which passes throw four steps (Network, Remote Sensing, Research, Showcase and awareness).
1.1. Permanent Sampling Plots Network
PSP or permanent sampling plots which are established on the ground in order to provide a framework for understanding forest dynamics and the possibility of linking Copernicus satellite data in monitoring and taking decision in the forest management process.
Fig. 1 – Distributions of permanent sampling plots on altitudes
In figure 1 is shown the distribution of the permanent sampling plots on different altitudes in order to cover the diverse forest structures at different ages
1.2. Field data methodology
One plot is a 1-hectare rectangular plot (100x100 m) in which the species and dbh (diameter at breast height) of every tree greater than 10 cm are measured and recorded (fig. 2).
Fig. 2 – Marteloscope site
Each plot was scanned with a TLS in order to extract: the location of each trees (relative position accuracy <2cm) and the 3D characteristics of each tree (volume, shape of the trunk, crown, branches (fig. 3).
Fig. 3 – FARO TLS
Each plot contains 16 grid cells marked in the field (fig. 4)
Fig. 4 – Marteloscope site field markings
1.3. Field work results
The 3D point cloud collected from the field scan is processed with VirtSilv and displayed online (fig. 5, 6)
Fig. 5 – VirtSilv from-above view of individual trees
Fig. 6 – VirtSilv 3D view of individual trees
1.4. Drone mapping
The mosaics will be used in for texture analysis and correcting any geo-location errors between the field data and the Copernicus data given the orthorectification and ground control points accuracies.
Fig. 7 – Forest drone view
Fig. 8 – Forest drone view
Fig. 9 – Sentinel2 satellite view
Fig. 10 – Sentinel2 satellite view
Fig. 11 – Sentinel2 satellite view
Fig. 12 – Process methodology
2.1 Data entry
For this step are using Sentinel-2 MSI: Multispectral Instrument, Level-1C, The Sentinel-2 data is provided by EU/ESA/Copernicus. The call for Sentinel 2 as data entry is done through the import’s widget, which gets the entire history of data and the freshly ingested data from the provider, ESA.
Fig. 13 – Sentinel-2 MSI
2.2. Producing and adjusting existing vegetation indices
Fig. 14 – Satellite image process map
A) Masking the clouds
We use the specific classification of clouds and cirrus to create the mask. Every band of Sentinel is then cleaned pixel by pixel to extract the clouds.
Fig. 15 – Satellite image with clouds
B) Merging the bands
We get the pixels not marked as clouds and we merge them into a median pixel. Based on the period we filter the Sentinel collection and we create the cloud free satellite image using merge bands and masking clouds.
Fig. 16 – Satellite image without clouds
C) Vegetation indices
We determine more vegetation indices: NDVI, msAVI, AVI, BI, SI.
Fig. 17 – Vegetation indices differences
Based on the multiple correlation between the field data and Copernicus data we will use a Machine Learning Algorithm to produce different thematic maps (e.g. Map of Forest development Phase, Map of Main Tree Species etc.) suitable for Forest Management.
Fig. 18 – Thematic map view of Brasov City
- Portal for increasing awareness
We built a portal where we are publishing the Copernicus results for forest managers. This portal is a simple tool, adapted for mobile, in order to be used by forest managers for consulting the Copernicus results developed during the project implementation.
Fig. 19 – Portal view which help in decision making for forestry management
- Future development
Steps to take:
- Finishing the field work to complete the network;
- Producing VHR orthomosaic with drone over the field network;
- Continue research on identifying ways of improving Copernicus data for forest management;
- Using the field network to explain the possibilities of using Copernicus data to forest managers to increase the awareness.
Fig. 20 – Road map for future development